The Power of Marriage: The Causal Effect of Parental Marital Status on Child’s Earnings Bob Wen Economics Department, Clemson University July 31, 2020 Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 1 / 35
Outline This presentation is the empirical part of my research about the causal effects of parental marital status on the child’s earnings. � Motivation. � Research questions and hypotheses. � Identification challenges and specification strategy. � Descriptive statistics. � OLS regression and post-estimation analysis. � Endogeneity and sample selection. � Panel data regression. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 2 / 35
Introduction Motivation Child’s Earnings by Parental Marital Status Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 3 / 35
Introduction Motivation Child’s Earnings-Work Experience Profile by Parental Marital Status Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 4 / 35
Introduction Research Questions and Hypotheses Research Question and Hypotheses � Question: Can parental marital status during childhood explain the child’s adult earnings after controlling for other factors? � Parental marriage effect on child’s earnings hypotheses: A stable marital relationship has a positive and significant effect on the child’s adult earnings. The influence goes through three channels: the “investment in child’s education” channel, the “intergenerational marriage persistence” channel, and the unobserved “endowment transmission” channel. The parental marriage effect interacts with parental family income and parental education. The parental marriage effect on child’s earnings is stronger when parental income is higher or when the child comes from a highly educated family. It is higher for sons than for daughters. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 5 / 35
Empirical Analysis Identification Challenge and Specification Strategies Identification Challenges Two endogeneity issues that may lead to inconsistent OLS estimates: � Omitted parental variable bias. How to disentangle the effect of parental marital status from other parental factors. � Endogenous sample selection. How to deal with endogenous sample selection due to child’s labour force participation choice. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 6 / 35
Empirical Analysis Identification Challenge and Specification Strategies Specification Strategies � Add relevant parental variables: Include parental family income and parental educational attainment in the model to rule out the parental income and education effects. � Sample selection bias correction: Take into account the child’s decision of participating in the labour market by running a selection probit model and then use the predicted probability of LFP or the IMR as an additional regressor. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 7 / 35
Empirical Analysis Total Effect and Direct Effect Total Effect and Direct Effect: A Simplified Path Diagram of SEM Child’s education attainment and child’s marital status are the endogenous mediator variables that help to explain the mechanism through which parental marital status affects child’s earnings. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 8 / 35
Empirical Analysis Total Effect and Direct Effect Regression Framework and Variables Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 9 / 35
Empirical Analysis Total Effect and Direct Effect Descriptive Statistics: Table Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 10 / 35
Empirical Analysis Total Effect and Direct Effect Descriptive Statistics: Figures Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 11 / 35
Empirical Analysis Total Effect and Direct Effect Descriptive Statistics: Figures (a) (b) (c) (d) Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 12 / 35
Empirical Analysis SEM Structural Equation Modelling (SEM) Model 1: Direct effect model ln ( child earnings ) = β 0 + β 1 parent marital status + β 2 ln ( parent family income ) + β 3 parent education + β 4 child schooling + β 5 child marital status (1) + β 6 child experience + β 7 child experience 2 + β 8 child gender + β 9 child region + ε Model 2: “Investment in child’s education” channel model child schooling = γ 0 + γ 1 parent marital status + γ 2 ln ( parent family income ) + γ 3 parent education + γ 4 child experience + γ 5 child experience 2 (2) + γ 6 child gender + γ 7 child region + u Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 13 / 35
Empirical Analysis SEM Structural Equation Modelling (SEM) Model 3: “Intergenerational marriage persistence” channel model child marital status = λ 0 + λ 1 parent marital status + λ 2 ln ( parent family income ) + λ 3 parent education + λ 4 child experience + λ 5 child experience 2 (3) + λ 6 child gender + λ 7 child region + v Model 4: Total effect model ln ( child earnings ) = α 0 + α 1 parent marital status + α 2 ln ( parent family income ) + α 3 parent education + α 4 child experience + α 5 child experience 2 (4) + α 6 child gender + α 7 child region + ε Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 14 / 35
Empirical Analysis SEM Direct Parental Marital Effect Model After controlling for the two parental factors and the two channels (child’s education and marital status), the direct effect of parental marriage on child’s earnings is positive but not significant. (3.5%) Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 15 / 35
Empirical Analysis SEM The “Investment in Child’s Education” Channel The parental marriage has a positive and significant effect on child’s educational attainment. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 16 / 35
Empirical Analysis SEM The “Marriage Attitude Transmission” Channel The parental marriage has a positive and significant effect on child’s decision of marriage. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 17 / 35
Empirical Analysis SEM Total Parental Marital Effect Model After controlling for parental family income and parental education, the parental marriage has a positive and significant effect on child’s adult earnings. (11.1% or more precisely 11.7%) Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 18 / 35
Empirical Analysis SEM GSEM model result Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 19 / 35
Empirical Analysis SEM SEM Models Result Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 20 / 35
Empirical Analysis SEM The Decomposition of Total Parental Marriage Effects on Child’s Earnings It can be done manually or using sem and gsem. The total effect of a successful parental marriage on child’s earnings is 0.111, meaning that the workers who grew up in homes in which their parents remained married earn 11.1% (or precisely 11.7%) more than their counterparts who were raised by divorced or separated parents, holding other factors constant. This total effect can be decomposed into the direct effect and the indirect effect. The former is the effect after controlling for both the “investment in child’s education” channel and the “intergenerational marriage persistence” channel. It is positive but not significant. The latter is the effect through the two channels, which is 0.076. The percentage of the total effect that is mediated through the two intergenerational transmission channels is 0.076/0.111=68.5%. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 21 / 35
Empirical Analysis Postestimation Analysis Properties of the Parental Marriage Effect on Child’s Earnings The parental marriage effect and the parental family income effect reinforce with each other. The earnings gap between the two “parental marital” groups is larger and significant for the workers from higher parental income families. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 22 / 35
Empirical Analysis Postestimation Analysis Properties of the Parental Marriage Effect on Child’s Earnings The parental marriage effect on child’s earnings increases with parental family income. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 23 / 35
Empirical Analysis Postestimation Analysis Properties of the Parental Marriage Effect on Child’s Earnings The earnings gap between the two parental marital groups is larger and significant for workers who have highly educated parents. Bob Wen (shihaow@clemson.edu) Economics, Clemson University Stata Conference 2020 24 / 35
Recommend
More recommend
